Use case
Coding agents: autocomplete, edits and code review
Tools like Cursor, Cline and opencode already speak the OpenAI or Anthropic protocol. The workflow is a cheap model for routine edits and a larger one for hard reviews.
Updated
Two speeds, one key
Autocomplete and simple edits want speed and a low per-token price. A deep review of a diff wants a stronger profile.
Because the model is chosen per request, the client can mix profiles without a second provider.
Isolate the editor key
The key lives on the developer machine. Issue a dedicated key with a spend limit and expiry so a leak cannot touch production keys.
Revoke a single editor key without affecting teammates: their keys are independent.
Watch the loops
Agent loops can retry and escalate. A spend limit is a hard ceiling against a runaway run.
In /account/logs you can see the model, token counts and applied price of every agent request, which makes per-seat cost observable.
Published model prices
USD per 1M tokens. The numbers come from the public catalog on every page rebuild — this is the sale price, not an estimate.
| Model | Input | Output | Cache |
|---|---|---|---|
| Claude Opus 5.5 | $2.70 | $8.00 | $0.14 |
| DeepSeek Flash | $0.50 | $2.00 | $0.01 |
| Claude Sonnet 5.5 | $1.33 | $6.67 | $0.07 |
| DeepSeek V4 Pro | $2.20 | $6.60 | $0.07 |
Top-up from $10, balance in USD. 1 USD = 1,000,000 UT. No subscription and no auto-charge.
Questions
Does Cursor work with LLMcode?
Yes: override the OpenAI base URL to https://llmcode.io/v1 and use an LLMcode key. See /integrations/cursor.
Does Claude Code work?
Yes: set ANTHROPIC_BASE_URL to https://llmcode.io and ANTHROPIC_AUTH_TOKEN to your key.
Which profile for autocomplete?
The fastest cheap profile. Check /models for the entry tier and its price.
How do I stop an expensive run?
Set a spend limit in UT on the key before the run. Requests above the balance return 402.